Role of Ethanol Plants in Dakotas Land Use Change: Incorporating Flexible Trends in the Difference-in-Difference Framework with Remotely-Sensed Data

Supplementary Information Data Notes-Soil Quality USDA-NRCS's soil surveys are conducted at pre-designated spatial units, known as map units (MUs), which represent common management requirements towards various land uses (Soil Data Viewer 6.0 User Guide, 2011 pp. 11). Although MUs are the finest spatial resolution in the soil surveys, they are composed of multiple map unit components that are horizontal strips of similar soil characteristics. The MUs may vary in size (2 acres to 2,000 acres) depending upon the variability among their respective map unit components. We aggregate LCC and slope up to the MUs using a 'Soil Data Viewer' application. The aggregation criteria are differ as LCC is a categorical variable and slope is a continuous variable. Representative slope was aggregated as a weighted-average of representative slope for all map unit components within each MU, where weights are the respective area-shares. Variable LCC was aggregated by a 'dominant condition' criterion that assigned the LCC value of the map unit component that was designated the highest area-share among other components. Note that the 'dominant condition' aggregation criterion may assign the LCC value that represents as little as 25% area-share. Higher LCC value was assigned when different LCCs had equal area-shares. Although the tie-breaker was applicable to only 4 out of 156 MUs in North Dakota (0.7% of the state's area), and only 2 out of 260 MUs in South Dakota (0.6% of the state's area). The DID model in conjugation with PSM This section discusses the working of a standard Difference-inDifference model (DID) in conjunction with Propensity Score Matching (PSM). We follow the DID model framework of 2 Abadie (2005). Consider a representative land parcel i with , i t C and , i t CS as its corn acreage and combined corn and soy acreage respectively at time period t. We introduce binary variables i d and t δ to designate treatment/control groups and pre-/post-treatment periods respectively. So, 1 i d = for treated parcels and 0 otherwise, and 1 t δ = for the years after an ethanol plant was established and 0 otherwise. Further, denote () t t + − as the set of post-treatment (pre-treatment) time periods with 0 t as the treatment year. 1 Intuitively, to evaluate a treatment effect for treated parcel i we would compare its corn acreage with and without the ethanol plant in the post-treatment era (i.e., , i i C 2 …